search

TRIQS / ctseg

A segment picture impurity solver with spin-spin interactions.
https://triqs.github.io/ctseg
Other
7 stars 3 forks source link

add test for 2 orbital dimer (currently failing do not merge!) #1

Closed the-hampel closed 5 months ago

the-hampel commented 2 years ago

Hi @nkavokine @parcollet, I tried to run a basic multiorbital, static interaction example and run into segmentation faults. This example is part of the triqs benchmarks (https://github.com/TRIQS/benchmarks/tree/master/Dimer) and runs perfectly fine with cthyb with the delta_interface, and with the old ctseg. I modified the example to use only density-density type interactions, and removed the off-diag onsite hoppings.

I created a backtrace and get slightly different backtraces for a run with Hartree_shift on (local levels) or off:

 frame #0: 0x00007fffcbd52495 solver_core.cpython-310-x86_64-linux-gnu.so`triqs::det_manip::det_manip<delta_block_adaptor>::get_x(this=0x0000000000cd9bc0, i=1154334476675265717) const at det_manip.hpp:347:61                                     
    frame #1: 0x00007fffcbd6c170 solver_core.cpython-310-x86_64-linux-gnu.so`long det_lower_bound_x<triqs::det_manip::det_manip<delta_block_adaptor>, tau_t>(this=0x00007fffffff12c8, i=1154334476675265717)::'lambda'(long)::operator()(long) const at util.hpp:24:46                                                     
    frame #2: 0x00007fffcbd6c0a8 solver_core.cpython-310-x86_64-linux-gnu.so`long lower_bound<long det_lower_bound_x<triqs::det_manip::det_manip<delta_block_adaptor>, tau_t>(triqs::det_manip::det_manip<delta_block_adaptor> const&, tau_t const&)::'lambda'(long), tau_t>(f=(unnamed class) @ 0x00007fffffff12c8, N=2308

Full backtraces: lldb_bt.txt lldb_bt_hartree_shift.txt

However, both seem to be related to some wrong dimensions handed over to detmanip. Before I start digging I thought I ask you whether you have an idea. Something must go wrong in util.hpp when calling something in detmanip. The value handed over seems to be very wrong 1154334476675265717 .

The branch I created runs automatically a test that reproduces the error.

the-hampel commented 5 months ago

Fixed and added by 746100500a90499b8aa70f84f9f5af1a4af2fc4e